Top critical review
One person found this helpful
Zombies? No thanks
on 28 August 2016
In this book David Chalmers presents a case against materialist reductionist accounts of consciousness, and in favour of a non-interactionist dualist account. I don’t think his case stands up.
Chalmers bases his theory very heavily on the conceivability – and therefore, in his view, the logical possibility - of zombies. Chalmers also believes in epistemic asymmetry – that one needs to have experienced consciousness to understand what it is and believe in it. These two positions are incompatible. If there can be zombies, then there can be a zombie version of Chalmers himself. Zombie Chalmers is functionally identical to conscious Chalmers. Therefore, just like conscious Chalmers, zombie Chalmers writes a book in which he expounds his belief in consciousness – the very book that I am currently reviewing. But the argument from epistemic asymmetry says that since zombie Chalmers has no experience of consciousness, he cannot have a belief in consciousness, and therefore cannot write any such book. So zombie Chalmers both does and does not write this book: a contradiction.
Chalmers cannot have both zombies and epistemic asymmetry: one of them must go. My view is that the zombies must go. Could there be anything more absurd than the idea of a zombie philosopher writing a book about consciousness? What could possibly cause the neural activity to occur in the zombie philosopher’s brain that results in the writing of such a book? I don't see how there could be a zombie Chalmers, and if there can be no zombie Chalmers, there can be no zombie anyone else. And if the zombies must go, then so, it would appear, must Chalmers’ non-interactionist version of dualism.
Why does Chalmers think zombies are possible? Ultimately it seems to be because of his belief in the widely accepted principle of causal closure at the physical level, i.e. that where there is sufficient cause for a physical event, there is always sufficient physical cause. However, I think Chalmers may not be applying the principle correctly. (He doesn't discuss it overtly, so I am having to infer.) Suppose we have a physical event A, and that A causes an event B which, after quantum wave collapse, can occur in two ways, either as B1 or as B2. Suppose that we actually get the sequence A -> B1. Is A a sufficient cause for B1? It depends what you mean by 'sufficient'. If you mean that B1 can follow A without the need for any additional event, then (according to quantum theory) yes, it is sufficient. But if you mean that A is enough to ensure that B1 occurs, then no, A is not sufficient. The question here is whether a full explanation of human behaviour requires only the first kind of sufficiency, or, unusually for physical processes, requires the second. Chalmers seems to be assuming that it only requires the first, but this is only an assumption. The fact is that we don't know (yet); and since human behaviour is atypical in many ways for physical processes - it is purposive, for a start, which most physical processes are not - and since Chalmers' assumption leads to the apparent absurdity of zombie philosophers writing books about consciousness when they have no consciousness to write books about, I think we would be justified in at least holding off assent to this assumption until it can be shown that it is correct.
Later in the book, Chalmers considers whether an interactionist form of dualism might fare better than his version. His remarks here seem to me rather superficial and, in some cases, downright odd. He notes that:
"… some have appealed to the existence of quantum indeterminacy, and have suggested that a nonphysical consciousness might be responsible for filling the resultant causal gaps…"
which, as I've already hinted, does seem to offer a possible solution to the problem of how phenomenal consciousness might be able to act causally on behaviour. But he dismisses the suggestion for reasons which, to me at least, seem ill-founded:
"First, the theory contradicts the quantum-mechanical postulate that these microscopic “decisions” are entirely random…"
In fact it does not, because as Chalmers himself has just noted, consciousness is presumed in this scenario to be nonphysical, so it remains the case that events are random at the quantum level insofar as physical processes are concerned.
"… and in principle it implies that there should be some detectable pattern to them – a testable hypothesis."
It seems odd to describe the testability of the suggestion as a 'problem'. Surely it is an advantage?
"Second,… it needs to be the case that the behavior produced by these microscopic decisions is somehow different in kind than that produced by most other sets of decisions that might have been made by a purely random process…. This again is testable in principle… to hold that the random version would lead to unusually degraded behavior would be to make a bet at long odds."
Well, again, testability is hardly a disadvantage. And I know Chalmers is a mathematician, but he can't expect us to take the length of the odds on trust: some evidence of the calculations should be provided. And anyway, instead of discussing the theory, why not just do the tests?
Chalmers then has some remarks about the possible influence of consciousness in collapsing quantum wave functions, and says:
"…it seems that the kind of causal work consciousness performs here is quite different from the kind required to play a role in directing behavior."
Is it? It seems to me to be essentially the same work. Both involve collapsing wave functions: the difference is that in one case the collapse is being directed, and in the other it is allowed to be random; but this is a difference within consciousness, not in the causal mechanism, whatever that may be, that constitutes the interaction between consciousness and the physical world.
He then goes on:
"In any case, all versions of interactionist dualism have a conceptual problem… We can always subtract the phenomenal component from any explanatory account, yielding a purely causal account."
Ah, but can we? Here is Chalmers earlier in the book, introducing the subject of zombies:
"To fix ideas, we can imagine that right now I am gazing out of the window, experiencing some nice green sensations from seeing the trees outside, having pleasant taste experiences through munching on a chocolate bar, and feeling a dull aching sensation in my right shoulder…. What is going on in my zombie twin? He is physically identical to me, and we may well suppose that he is embedded in an identical environment. He will certainly be identical to me functionally…"
Stop right there. ‘Identical to me functionally’? This would mean that if, for example, you enter the room where zombie Chalmers is sitting, he might do any or all of the following things that Chalmers himself might do (I don’t know Chalmers, and for all I know he wouldn’t do any of these things; but there will be things he might do in this situation, and anyone who knows him is free to substitute those – my argument will still go through):
- point to the tree outside and say, ‘Look – aren’t those leaves a delightful fresh green colour?’
- offer you a chocolate bar like the one he is eating and say, ‘Here – have one of these, they’re delicious’
- rub his shoulder and say, ‘I really must put something on this – that fifth game of squash wasn’t such a good idea.’
But why would zombie Chalmers do any of these things, since he can’t experience colours, tastes or dull aches? He has no motivation to do them; so he won’t do them. And yet, as a philosophical zombie, he is functionally identical to the conscious Chalmers, and if they are what the conscious Chalmers would do, then he will do them. So he will do them, and he will not do them: another contradiction. This contradiction arose because we supposed that we could ‘subtract the phenomenal component from any explanatory account’. It turns out that we can’t – not without contradicting ourselves.
The point here is that phenomenal consciousness is causally explanatory at the level of human behaviour. We therefore cannot merely subtract it and leave the causality intact. Whether this should make us interactionists is not clear to me; but I think it shows, again, that Chalmers' non-interactionist version of dualism must be wrong.
A further reason to reject Chalmers’ theory is that his account of the supposed connection between conscious experiences and their physical correlates is seriously flawed. This will take a little while to explain.
Chalmers has some remarks about the ‘psychophysical laws’ which he suggests regularise the correlation of particular conscious experiences with particular physical events in the body:
"There is a system of laws that ensures that a given physical configuration will be accompanied by a given experience, just as there are laws that dictate that a given physical object will gravitationally affect others in a certain way.
"It might be objected that this does not tell us what the connection is, or how a physical configuration gives rise to experience. But the search for a connection is misguided. Even with fundamental physical laws, we cannot find a “connection” that does the work. Things simply happen in accordance with the law; beyond a certain point, there is no asking “how”… If there are indeed such connections, they are entirely mysterious in both the physical and psychophysical cases, so the latter poses no special problem here.
"It is notable that Newton’s opponents made a similar objection to his theory of gravitation: How does one body exert a force on another far away? But the force of the question dissolved over time. We have learned to live with taking certain things as fundamental."
Well, let’s consider this by way of an example. Suppose I cut my finger. This causes a message to be sent along the nerves to my brain, where two things happen: one, I feel a pain that appears to me to be localised to the injured area; and two, messages are sent from my brain to other parts of my body to initiate appropriate physical action (such as saying ‘Ouch’ and sucking my injured finger). According to Chalmers, the pain I feel has no causal role, and I would say ‘ouch’ and suck my finger even if I were a zombie and felt no pain at all. The fact that I feel pain is the result of the coming into play of a psychophysical law that ensures that, even though the pain does nothing, this is nevertheless what happens whenever I cut my finger.
There are three things to say here. Firstly, we do actually know what happens when people cut themselves and feel no pain, because there are such people: they suffer from a condition known as ‘congenital analgesia’, and their inability to feel pain, far from making them behave exactly like those of us who do feel pain, means that when something happens to their bodies that would cause us to take some appropriate action, they do nothing, with the result that they have an increased risk of further damage, e.g. from infection. So the suggestion that a being like ourselves but without consciousness would behave as we do when experiencing pain runs counter to our actual experience.
Secondly, the parallel with the law of gravity is highly misleading. As I pointed out, when I cut my finger, on Chalmers’ theory two distinct things occur: a feeling of localised pain, and the initiation of physical action such as saying ‘ouch’ and sucking the injured finger. In the case of gravity, only one thing occurs – an attraction between two physical bodies. In the case of pain, there is a conceptual similarity between the pain and the action of sucking the injured finger: they are both localised, one subjectively and the other objectively, to the same location on the body. No such conceptual similarity obtains in the case of gravity, because there are not two things in that case but only one.
Thirdly, it seems an astonishing coincidence that, despite the fact (according to Chalmers’ theory) that subjective experience is not causal, a psychophysical law nevertheless obtains which puts in place a subjective experience that is precisely what would be required if it were causal. To see how acute this problem is, consider how unlikely it would be that, if we chose a subjective experience at random to occur when I cut my finger, it was a localised feeling of pain in just that injured finger. It could, after all, be a pain localised to some other part of my body, or the smell of strawberries, or the sound of a trumpet. Yet inexplicably - literally inexplicably, according to Chalmers - there is a fundamental law that puts in place precisely the subjective experience needed to cause me to attend to my injured finger even though it doesn’t do this at all. This is just too odd and suspicious to accept. The world does not work like this: it is not founded on astronomically unlikely coincidences for which there is no explanation. Chalmers here is asking us to believe something quite unbelievable.
To approach the same point from the opposite direction: If we suppose for a moment, contra Chalmers, that phenomenal experience is causal, then what experience would best suit the purpose of directing my attention to the injured finger so that I might take some soothing or healing or protective action? The answer seems obvious: it should be an unpleasant feeling (so that I am impelled to address whatever it is that is causing the unpleasantness, in this case the untreated injury), it should contain no extraneous subjective content that might confuse the issue and impel me to take some irrelevant action (for instance, it should not include the smell of strawberries or the sound of a trumpet, since these would be entirely unhelpful), and it should be subjectively localised to precisely the place on my body that requires attention, so as to direct my attention to the right spot. And all three of these features, amazingly – amazingly, that is, if the phenomenal pain is not causal - are possessed by what I do in fact experience when I cut my finger.
It is here, I think, that Chalmers' non-interactionist dualism runs into its worst difficulties. If the physical world is truly causally closed, as Chalmers believes, then phenomenal experience is causally irrelevant. In that case, phenomenal experience can be configured in any way we like to imagine, without having the slightest effect on human behaviour. It would then seem that there are a vast number of possible worlds, each with its own peculiar set of psychophysical laws, in which there are counterparts to ourselves who behave exactly as we do, but have very different phenomenal experiences. In one of these worlds, our counterparts hear only the sound of trumpets, irrespective of what they are doing at the time. In another, they merely get smells - of onions, say, or strawberries. In only a tiny subset of these worlds do our counterparts have phenomenal experiences which are such that they meet the three criteria I mentioned above, criteria which make it seem plausible that they could be causes of behaviour rather than merely accompanying it. And now the non-interactionist has a problem, which is to explain how it comes about that we happen to live in one of this tiny subset of worlds, rather than in one of the others. Given the prima facie plausibility of the suggestion that phenomenal experience may be causal after all, it is clearly not enough to shrug this off and say, 'Well, we just do': the odds against it are so enormous that such an attitude is obviously unreasonable. The only possible avenue that I can see for the non-interactionist is to claim that this close correlation between phenomenal experiences and physical events is a feature of all possible worlds - in other words, that it is metaphysically necessary. Yet it is hard to see how a defence of this position could be mounted. The correlation certainly does not seem to be logically necessary, and it is hard to see how to show metaphysical necessity by any other route.
If, however, conscious experience is causal, everything becomes much easier to explain. The reason why, when I cut my finger, I have a subjective experience that is so precisely what is required to direct my attention to my injured finger, is that this has survival value, whereas the smell of strawberries, or the sound of a trumpet, or a pain elsewhere in my body, would not – in fact their survival value would be negative. It seems obvious that the reason so few people suffer from congenital analgesia is that the inability to feel phenomenal pain confers a survival disadvantage, and therefore is selected against by evolution; and this can only be the case if phenomenal pain is causally efficacious. If we allow conscious experiences to be causal, then it is evolutionary advantage that has given us the particular conscious experiences we have. This seems to me a much more plausible position than one which asks us to accept that there is an unexplained and highly improbable law that creates these correlations.
The rest of the book pretty much degenerates into mere metaphysical speculation. Chalmers speculates that organizational complexity plus some unknown psychophysical laws may be enough to create consciousness, and on this basis he invites us to ask ourselves what the conscious life of a thermostat may be like, and whether if you got enough Chinese people talking to each other with radios this might generate a group phenomenal consciousness. Ironically, Chalmers, who invented the term 'hard problem', is here forgetting just why the problem is hard: it is hard precisely because, whatever happens in the physical world and however we describe it, we are unable to see why phenomenal consciousness should be the result. The suggestion that organization may be a key to the problem is really just another kind of reductionism. The fact is that we simply don't know why or how brains manage to generate consciousness, and there is really no point building castles in the air on foundations of mere metaphysical speculation.
At the end of this book I am left with the strong feeling that, while trying to solve the mind-body problem using the tools of philosophy is an entertaining parlour game, no real progress will be made until science steps in and does the job properly. Which, of course, may never happen.